EPHA2/CD44-directed trafficking enhances endosomal leakiness and antisense therapy delivery

Antisense oligonucleotides (ASOs) are short single-stranded DNA molecules designed to target complementary mRNA and regulate its expression. Binding of an ASO to mRNA forms RNA–DNA hybrids that are recognized and degraded by the endonuclease RNase H1 (Cerritelli and Crouch, 2009; Nowotny et al., 2007), leading to reduced synthesis of the encoded protein. RNase H1 localizes to the nucleus and mitochondrion, where nascent RNAs are synthesized (Liang et al., 2017), but can also act in the cytoplasm where most mature transcripts reside. Thus, access to the appropriate cellular compartment is critical for ASO activity. To facilitate delivery, formulations such as liposomal encapsulation or sugar conjugation (Roberts et al., 2020) have been used; however, these approaches can cause unwanted side effects, including aberrant immune responses (Coelho et al., 2013; Moschos et al., 2007).

A newer generation of constrained ethyl-ASOs (cET-ASOs) enables vehicle-free delivery (Byrnes et al., 2023; Stein et al., 2010), thought to occur via endocytosis. Following uptake, cET-ASOs are sorted within the endosomal network, and their ability to escape from endosomal lumens into the cytoplasm or nucleus determines their capacity to engage target mRNAs and recruit RNase H1 (Linnane et al., 2019). Two major limitations are recycling to the extracellular space and trafficking to degradative compartments such as lysosomes, both of which reduce opportunities for endosomal escape (Kapustin et al., 2021; Linnane et al., 2019). Despite the importance of endocytic trafficking for ASO efficacy, the mechanisms governing their uptake and intracellular routing remain poorly understood.

We recently described a mechanism in which endosomes are directed to the nuclear vicinity and associate with nuclear pores (Marco et al., 2021). This “nuclear-capture” process is mediated by the cytoplasmic tail of the receptor tyrosine kinase EPHA2 (often overexpressed in tumors [Wykosky and Debinski, 2008]), through its interaction with the nuclear import machinery. Nuclear capture is mediated by phosphorylation of EPHA2 at Ser⁸⁹⁷, followed by endocytosis and positioning of EPHA2-containing endosomes near the nucleus.

High EPHA2 expression is often seen in tumor types which are driven by KRAS-activating mutations (Dunne et al., 2016; Markosyan et al., 2019; Mudali et al., 2006; Thaker et al., 2004). In pancreatic cancer, KRAS mutation is a key driver of tumor initiation, and mutant Kras deletion induces regression in preclinical models (Collins et al., 2012; Ying et al., 2012). We therefore investigated whether an EPHA2-dependent endocytic route leading to nuclear capture could be exploited to enhance uptake of a KRAS-targeting cET-ASO (cET-ASO^Kras) in cells from pancreatic cancers.

In human pancreatic ductal adenocarcinoma (PDAC), EPHA2 is strongly expressed in tumor nodules but largely absent in adjacent normal tissue (Fig. 1, A and B; and Fig. S1 A). Key PDAC drivers—mutated KRAS, TP53, CDKN2A, and SMAD4—correlate with increased EPHA2 (Fig. S1, B–F), and EPHA2 levels associate with reduced overall survival and the aggressive squamous subtype (Fig. S1, G and H). Notably, human PDACs exhibit EPHA2 Ser⁸⁹⁷ phosphorylation (Fig. 1, A and B), a requirement for nuclear capture of endosomes (Marco et al., 2021). Similarly, tumors from the KPC mouse (Kras^G12D/+; LSL-Trp53^R172H/+; Pdx1-Cre), which mimics metastatic squamous PDAC (Gopinathan et al., 2015), express abundant phospho-Ser⁸⁹⁷ EPHA2 (Fig. 1 C). To test KRAS dependency for PDAC growth, KPC-derived spheroids were treated with the Kras-targeting cET-ASO, cET-ASO^Kras. cET-ASO^Kras was taken up by KPC cells, and this significantly reduced spheroid volume and cell number (Fig. 1, D–G), confirming that ASO-mediated Kras inhibition can oppose PDAC growth in 3D culture.

To investigate EPHA2’s role in cET-ASO^Kras uptake into PDAC, we used tumor cell lines from KPC mice that were either WT (KPC-Epha2^+/+) or knockout (KPC-Epha2^−/−) for Epha2. In KPC-Epha2^+/+ cells, cET-ASO^Kras reduced Kras mRNA dose dependently, reaching 50% inhibition at ∼0.1 μM (Fig. 1 H). Contrastingly, in KPC-Epha2^−/− cells, cET-ASO^Kras’s ability to suppress Kras expression was reduced by >100-fold (Fig. 1 H). Western blotting confirmed that 1 μM cET-ASO^Kras suppressed KRAS protein levels and downstream MEK/ERK phosphorylation in KPC-Epha2^+/+, but not KPC-Epha2^−/− cells (Fig. 1 I). Accordingly, cET-ASO^Kras inhibited growth only in KPC-Epha2^+/+ cells (Fig. 1 J). Importantly, a nontargeting ASO (cET-ASO^NT) affected neither Kras expression nor MEK/ERK signalling (Fig. S1, I and J). EPHA2’s necessity for ASO-mediated Kras targeting was further validated by CRISPR-mediated Epha2 deletion in KPC (Fig. S1, K–M) and human H1299 lung carcinoma cells (Fig. S1, N–P).

We hypothesized that EPHA2 facilitates cET-ASO^Kras endocytosis and subsequent nuclear capture of ASO-containing endosomes to suppress Kras expression. Indeed, endocytosed cET-ASO^Kras localized to nuclear-proximal, EPHA2-positive vesicles 4 h after addition (Fig. 2, A and B), and ASO internalization was significantly reduced in Epha2 knockout cells (Fig. 2 C), confirming EPHA2’s role in ASO endocytosis and trafficking.

We previously linked EPHA2-mediated nuclear capture to EphA2-Ser⁸⁹⁷ phosphorylation, Rab17-dependent trafficking, and NLS-driven nuclear capture. Investigating the requirement for these processes for productive ASO uptake, we report the following:

• 1

  EphA2 phosphorylation: cET-ASO^Kras increased EphA2-Ser⁸⁹⁷ phosphorylation (Fig. 2 D), and a phospho-defective EPHA2 mutant (EPHA2^S897A) failed to fully rescue ASO-driven Kras suppression in Epha2^−/− cells (Fig. 2 E).

• 2

  Rab17-dependent trafficking: EPHA2 knockout reduced ASO trafficking to RAB17-positive endosomes (Fig. 2 F). Consistently, cET-ASO^Kras’s ability to suppress Kras and its downstream signalling was compromised in RAB17 knockout cells (Fig. 2, G and H).

• 3

  Nuclear capture: While EPHA2^WT restored cET-ASO^Kras activity in Epha2^−/− cells, an NLS mutant (EPHA2^NLS)—which prevents nuclear capture (Marco et al., 2021)—did not (Fig. 2, I and J). Accordingly, cET-ASO^Kras suppressed growth in EPHA2^WT but not EPHA2^NLS cells (Fig. 2 K).

Finally, bafilomycinA1 or chloroquine reduced endosomal accumulation of ASO (Fig. 2 L and Fig. S1 Q), indicating endosomal acidification is required for trafficking to nuclear-proximal vesicles. Together, these data demonstrate that cET-ASO^Kras efficacy requires EPHA2-dependent endocytosis and Rab17-dependent trafficking to nuclear-captured endosomes.

Since cET-ASO increases EPHA2 Ser⁸⁹⁷ phosphorylation (Fig. 2 D), and phospho-defective EPHA2^S897A does not support cET-ASO^Kras uptake (Fig. 2 E), we investigated the signalling underlying ASO-driven EPHA2 phosphorylation. Various kinases, like AKT and p90RSK, phosphorylate Ser⁸⁹⁷ following growth factor or stress cues (Hamaoka et al., 2018; Harada et al., 2015; Miao et al., 2015; Zhou et al., 2015). Western blotting showed KPC cells rapidly activate p90RSK, but not AKT, in response to cET-ASO (Fig. 3 A and Fig. S2 A). MAPK signalling (MEK, ERK, and JNK, but not p38MAPK), which typically activates p90RSK, also increased (Fig. 3 A and Fig. S2 B). Consistently, a p90RSK inhibitor LHJ685 (RSKi) blocked cET-ASO–driven p90RSK activation (evidenced by reduced pSer^235/236-RPS6) and EPHA2 Ser⁸⁹⁷ phosphorylation (Fig. 3 B), and p90RSK inhibition decreased cET-ASO^Kras’s ability to suppress Kras levels (Fig. 3 C). Alongside this, we found that cET-ASO still activated ERK, JNK, and p90RSK in EPHA2 knockout KPC or H1299 cells (Fig. 3 D; and Fig. S2, C and D). Taken together, these data indicate that cET-ASO–driven activation of MAPK and p90RSK signalling is upstream of EPHA2-Ser⁸⁹⁷ phosphorylation.

We next investigated if a receptor-mediated mechanism drives cET-ASO–induced MAPK/p90RSK signalling and EPHA2 phosphorylation. Scavenger receptors (SRs) mediate myeloid antigen uptake and hepatocyte lipid clearance (Terpstra et al., 2000; Yu et al., 2015). SRs also internalize polyanionic molecules like ASOs (Miller et al., 2016; Tanowitz et al., 2017) and initiate MAPK signalling (Grewal et al., 2003; Yang et al., 2020a). Our screen of KPC (Fig. 3 E) and H1299 (Fig. S2 E) cells identified SCARB1 (SR-B1) and CD44 (SR-K1) as highly expressed in both cell types. CD44 and SCARB1 localized at or near the plasma membrane in human PDAC tumors (Fig. S2, F and G); notably, CD44 expression is highest in the most aggressive squamous PDAC subtype (Bailey et al., 2016).

As cET-ASO^Kras was endocytosed into perinuclear endosomes positive for EPHA2 and CD44 (Fig. 3, F and G), we considered whether CD44 and cET-ASOs directly associate to form a receptor–ligand complex. Fluorescence polarization spectroscopy showed that Cy3-labelled cET-ASO and the canonical CD44 ligand, hyaluronic acid (HA), both directly bind purified CD44 with micromolar affinity (of 10.8 and 1.8 µM, respectively) (Fig. 3 H), and unlabelled HA displaced Cy3-cET-ASO from CD44 (Fig. 3 I), suggesting they compete for the same site. Consistently, proximity ligation assays showed that cET-ASO^Kras and CD44 associate in situ (Fig. 3, J and K). To test if cET-ASOs can physically connect EPHA2 with CD44, we used an EPHA2-TurboID fusion to perform proximity-dependent biotinylation. cET-ASO^Kras addition increased CD44 biotinylation by EPHA2-TurboID (Fig. 3 L) and the appearance of biotin-labelled proteins in intracellular vesicles (Fig. 3 M). Together, these data indicate that cET-ASOs interact directly with SRs to promote physical association and co-endocytosis of an EPHA2–SR complex.

We next investigated how SRs influence cET-ASO internalization, MAPK/p90RSK signalling, and KRAS suppression. CRISPR-mediated deletion of Cd44 or Scarb1 in KPC cells (Fig. S2, H and I) blocked cET-ASO–driven activation of MAPKs and p90RSK, as well as EPHA2 Ser⁸⁹⁷ phosphorylation (Fig. 3 N and Fig. S2 K). Accordingly, vesicles containing ASOs and/or internalized EPHA2 were undetectable in Cd44-deficient cells (Fig. 3 O). Furthermore, deleting Cd44 or Scarb1 prevented cET-ASO^Kras from suppressing Kras expression in KPC (Fig. 3 P and Fig. S2 J) and H1299 cells (Fig. S2, L–N). These data indicate that cET-ASOs activate SRs to trigger p90RSK signalling, CD44-EPHA2 association, and EPHA2 Ser⁸⁹⁷ phosphorylation. This process then facilitates the co-trafficking of cET-ASO, SRs, and phospho-EPHA2 to juxta-nuclear endosomes, enabling productive ASO uptake and targeting Kras mRNAs.

Endosomal escape of ASOs to the cytoplasm, through mechanisms which remain largely unclear, is required for mRNA targeting. Reactive oxygen species (ROS) can damage internal membranes to increase their permeability and allow endosomal escape—a process which is thought to be linked to antigen cross-presentation (Bhardwaj et al., 2023; Canton et al., 2021; Dingjan et al., 2016). Reports that CD44 signalling is involved in both antigen presentation and ROS production in endo-lysosomal compartments (Niemietz and Brown, 2023; Vachon et al., 2006) further prompted us to investigate if ASO uptake evokes endomembrane damage. Treating KPC cells with cET-ASO^Kras significantly increased ROS levels, as measured by nuclear translocation of CellROX green (Fig. 4, A and B). To detect lipid peroxidation (LP)—a key driver of ROS-mediated membrane damage—we used the C11-Bodipy probe (Drummen et al., 2002). cET-ASO^Kras addition increased LP (Fig. 4, C and D) specifically within intracellular vesicles (Fig. 4 E). Moreover, the radical-trapping antioxidant liproxstatin-1 blocked these ASO-induced increases in both ROS and LP (Fig. 4, A and D).

While endosomal leakiness promotes ASO release, it also recruits repair effectors like ESCRT components or galectins (galectin-3, -8, and -9) (Gros et al., 2022; Hedlund et al., 2023; Jia et al., 2018; Jia et al., 2020; Skowyra et al., 2018; Wittrup et al., 2015). We detected galectin-9 as a proxy for leaky membranes and thus identified potential sites of cET-ASO endosomal escape. In KPC-Epha2^+/+, but not KPC-Epha2^−/−, cells, cET-ASOs increased the size and overlap of galectin-9/ASO-positive vesicles (Fig. 5, A and B; Fig. S3, A and B). BafilomycinA1 treatment reduced galectin-9 recruitment to ASO-positive vesicles (Fig. S3 C), suggesting endosomal maturation is required for ASO-induced leakiness. Finally, cells expressing nuclear capture–defective EPHA2 (EphA2^NLS) showed decreased colocalization of cET-ASO with EPHA2 and galectin-9 (Fig. 5, C and D). This reduction was most pronounced in nuclear-proximal vesicles (Fig. 5, E and F), indicating EPHA2 nuclear capture is essential for trafficking ASOs to leak-prone endosomes.

LP and endo-lysosomal damage drive liquid–liquid phase separation, triggering assembly of ribonucleoprotein and stress granules (SGs) (Balakrishnan and Kenworthy, 2024; Yang et al., 2020b). Moreover, G3BP1, a core SG component, has recently been shown to be recruited to leaky endosomes to facilitate their repair (Bussi et al., 2023). Following cET-ASO addition, many nuclear-proximal structures were triply positive for cET-ASO, galectin-9, and G3BP1 (Fig. S3 D). This G3BP1 recruitment was EPHA2-dependent; KPC Epha2^−/− cells largely failed to form G3BP1 condensates (Fig. 5, G and H), which, if present, were reduced in size and number (Fig. 5 I and Fig. S3, E–G). Furthermore, EPHA2, EPHA2^pS897, and CD44 overlapped with G3BP1/galectin-9 structures (Fig. S3, H and I), linking them to the nuclear-capture pathway. Collectively, these data indicate that cET-ASOs, via SR signalling, promote EPHA2^pS897-dependent trafficking to nuclear-proximal compartments. These compartments subsequently become leaky, enabling endosomal escape and productive cET-ASO^Kras uptake.

Cellular stresses activate kinases like PKR, PERK, GCN2, and HRI, increasing eIF2α phosphorylation and triggering SG assembly as part of the integrated stress response (ISR) (Klein et al., 2022). Addition of cET-ASO increased phospho-eIF2α in KPC cells (Fig. 5 J), but not in Epha2-or Cd44-knockout KPC cells. This suggests EPHA2-dependent cET-ASO internalization and delivery to nuclear-captured endosomes drives SG assembly. Fluorescence in situ hybridization using a poly-d(T) probe showed close colocalization of polyadenylated mRNAs and G3BP1 with internalized cET-ASOs (Fig. 5 K). Similarly, cET-ASO addition induced G3BP1/galectin-9–positive structures associated with the SG marker EIF3b (Fig. S3 J) (Jayabalan et al., 2016). Thus, cET-ASO–positive endosomes in EPHA2-expressing cells associate closely with bona fide SG markers.

Beyond regulating mRNA translation, SGs recently were shown to plug endo-lysosomal membrane holes caused by mycobacteria, restricting their leakage and pathogenicity (Bussi et al., 2023; Jia et al., 2022). We proposed that SG formation at damage sites similarly limits cET-ASO efficacy. To test this, we used ISRIB (ISR inhibitor), which opposes SG assembly by blocking eIF2α phosphorylation effects (Rabouw et al., 2019; Sidrauski et al., 2015). As expected, ISRIB inhibited cET-ASO–driven SG assembly and G3BP1 recruitment without affecting endosomal ASO uptake (Fig. 5, L–N; and Fig. S3, K and L). Notably, ISRIB enhanced the ability of cET-ASO^Kras to suppress Kras mRNA expression (Fig. 5 O). SG assembly can be opposed by combined knockout of its core components G3BP1 and G3BP2 (Yang et al., 2020b). CRISPR-mediated deletion of both genes (G3bp^{CRISPR dKO}) (Fig. S3 M) significantly enhanced cET-ASO^Kras efficacy in suppressing Kras expression in KPC cells (Fig. 5 P). These data indicate that EPHA2-dependent trafficking to leaky, nuclear-proximal endosomes is essential for efficient target gene suppression, and that ASO efficiency is enhanced by inhibiting SG-mediated endosomal repair mechanisms (Fig. S3 N).

The efficacy of genetic medicine can be limited by poor endosomal escape. We identified a pathway where cET-ASOs bind a SR, CD44, which cooperates with EPHA2 to drive ASO uptake and trafficking to nuclear-proximal endosomes. These nuclear-captured compartments become leaky, allowing ASO escape into the cytoplasm (Fig. S3 N). Pharmacological or genetic inhibition of SG-mediated membrane repair further enhances this efficacy. SR-dependent activation of p90RSK-driven EPHA2 phosphorylation initiates this process. While SRs traditionally mediate antigen uptake and endosomal leakage for MHC presentation in immune cells (Barth et al., 2008; Gros et al., 2022; Urban et al., 2001), in PDAC cells, SRs are more likely to support nutrient uptake and stemness (Guerrero-Rodríguez et al., 2022). Thus, we propose that a receptor which has been selected for its ability to support tumor stemness and growth may be exploited as a gatekeeper to an endocytic pathway capable of delivering therapeutic molecules to aggressive tumors. CD44 may promote endosomal leakiness by influencing iron uptake, either through Fe³⁺-loaded HA recruitment or TfR upregulation (Ando et al., 2025; Caneque et al., 2025). This Fe³⁺ can drive Fenton reaction-mediated LP, permeabilizing the ASO-loaded compartments (Saimoto et al., 2025). Thus, CD44 facilitates productive ASO uptake both by trafficking molecules to specific endosomes and by promoting the membrane damage necessary for their release.

While RTKs like EGFR are known to facilitate cET-ASO endocytosis (Wang et al., 2018), subsequent trafficking determines if ASOs can reach target mRNAs. Our evidence shows EPHA2 drives both cET-ASO uptake and trafficking to leaky endosomes. Crucially, Kras-mutant tumors, including most PDAC, express high levels of EPHA2 (Fig. S1 B), making them ideal candidates for cET-ASO therapy. Although EPHA2 normally functions during development by engaging ephrin ligands leading to autophosphorylation on tyrosine and trafficking to lysosomes, in cancer high levels of unengaged EPHA2 in cancer allow EPHA2-Ser⁸⁹⁷ phosphorylation in response to specific cues (Gundry et al., 2017; Marco et al., 2021). Rather than terminating signalling in lysosomes (Hiramoto-Yamaki et al., 2010), this phosphorylation enables nuclear capture of EPHA2-containing endosomes (Marco et al., 2021). The proximity of these leaky vesicles to the nucleus, governed by EPHA2’s NLS sequences, mechanistically links nuclear capture to endosomal leakiness.

Endosomal leakiness varies by compartment; early/recycling endosomes offer the highest probability for cargo escape (Paramasivam et al., 2022). Consequently, directing ASOs to these compartments can boost their activity (Finicle et al., 2023). Our data show cET-ASOs accumulate in RAB17, RAB11, RAB14, and RCP-positive recycling endosomes, linking the recycling characteristics of nuclear-captured vesicles with leakiness. Furthermore, blocking endosomal acidification/maturation—which prevents cargo delivery to recycling endosomes—compromised cET-ASO accumulation, confirming that this type of compartment facilitates cytosolic leakage.

We also identified SG recruitment to damaged endosomes occurring in response to ASO-induced leakiness. Traditionally known as mRNA translation regulators (Buchan and Parker, 2009; Moon et al., 2019), SGs are now implicated in endomembrane repair (Bussi et al., 2023; Jia et al., 2022). Our finding that chemical SG inhibition increases cET-ASO efficacy suggests that pharmacological tools targeting the translation machinery could be exploited to enhance the delivery of endocytosed therapeutics.

Immunofluorescence staining was performed on paraffin-embedded tissue sections of patient pancreatic adenocarcinoma (PAAD; PDAC) tumors or KPC mouse PDAC tumor sections. Human tissues were retrieved through patients undergoing surgery with curative intent within Greater Glasgow and Clyde NHS hospitals through the Glasgow Biorepository (N.526) (n = 7). KPC tissue samples were obtained from archived paraffin blocks used in previous studies (Gundry et al., 2017).

Briefly, paraffin was removed from tumor sections containing slides using sequential steps with xylene, 100, 95, and 80% ethanol washes and a final step with H₂O rehydration. Antigen retrieval was performed in citrate buffer, pH 6.0, and samples were placed in a water bath at 100°C for 20 min and allowed to cool down at RT for a further 20 min. Slides were washed three times with Tris-buffered saline (TBS), and samples were incubated with blocking buffer (1% normal goat serum and 0.1% Triton X-100 in TBS) for 1 h at RT. Primary antibodies were incubated ON at 4°C. The next day, slides were washed three times with TBS, and the corresponding secondary antibodies together with DAPI were incubated on each slide for 1 h at RT. Next, samples were washed three times with TBS buffer, and coverslips were mounted afterward using Fluoromount G mounting medium. Samples were visualized using a Zeiss LSM880 microscope with a 20×/0.8 NA objective and Zen Black 2.3 software; sequential excitation was used, and pixel size was 0.06 µm. Image analysis to measure the mean intensity of each fluorophore was performed across the different patient tissues in different fields of view for each patient using Fiji-ImageJ software (2.6.0).

KPC and H1299 human non-small cell lung carcinoma cell lines were cultured in DMEM, high glucose, and GlutaMAX, supplemented with 10% FBS (vol/vol). Cells were cultured at 37°C in a controlled humidified environment containing 5% CO₂, and cultures were split every 3–4 days using trypsin/EDTA solution. Cells were tested for the presence of mycoplasma, and H1299 cells were validated using the Promega GenePrint 10 system (STR multiplex assay) at the Molecular Technology Services (CRUK-Scotland Institute).

Agarose 9 × 9 micro-molds were prepared as per the manufacturer’s instructions (Microtissues 3D Petri Dish micro-mold, Merck) and placed in each well of a 12-well cell culture plate. KPC cells were trypsinized, and 4.5 × 10⁵ cells were seeded into each agarose micro-mold using 190 ul of medium. Cells were allowed to enter the micro-wells for 2 h at 37°C, and the medium was topped up immediately after covering the whole micro-mold. Cells were allowed to form spheroids for 3 days, and on DIV3, cET-ASO^Kras was added to each corresponding well. On DIV7, the spheroids were fixed while inside the micro-molds using 4% paraformaldehyde for 1 h at RT. Next, the spheroids were popped out from the microwells using PBS, collected, and transferred to an empty 96-well plate for manipulation. Spheroids were then incubated with blocking buffer (0.1% Triton X-100, 2% BSA, and PBS) for 30 min at RT. Primary antibodies were diluted in blocking buffer and incubated with the spheroids ON at 4°C. The next day, three 30-min washes with blocking buffer were performed for each set of spheroids, and then samples were incubated with the corresponding secondary antibodies as well as DAPI. Samples were incubated ON at 4°C. The next day, the spheroids were washed three times for a 30-min period each time, and finally samples were mounted with Fungi solution (60% glycerol, 2.5 M fructose, and H₂O) on a slide, using a frame seal (Bio-Rad) and a coverslip. Samples were visualized using either a Zeiss LSM880 confocal microscope as above or a Revvity Opera Phenix automated confocal microscope using a 20×/1.0NA water immersion objective (part HH14000421). Z-stacks using a 1 µm step were obtained for each spheroid and used to measure spheroid cell number and spheroid total volume. Data shown correspond to the individual spheroids pooled from independent experiments. Image analysis was performed using Harmony High-Content image analysis software (Revvity).

KPC or H1299 cells were seeded at a density of 1.5 × 10⁵ cells for each condition in a 6-cm plate. Cells were allowed to attach and proliferate overnight. The next day, cells were treated with either vehicle (PBS) or cET-ASOs at 0.05, 0.1, 1, 5, and 10 μM concentrations for 72 h. Next, RNA was harvested using Trizol (Thermo Fisher Scientific). 1 μg of RNA was used for RT-PCR to produce cDNA (High-capacity RNA to cDNA kit, Thermo Fisher Scientific) following the manufacturer’s recommended protocol. Kras mRNA expression was assessed using quantitative PCR (qPCR) using Quantinova SYBR Green RT-PCR Kit (Qiagen), a CFX1000 thermocycler (Bio-Rad), and CFX Maestro software (Bio-Rad). RPLP0 ribosomal RNA was used as a reference gene. Kras mRNA expression is expressed relative to untreated cells for each condition across independent experiments.

KPC or H1299 cells were seeded at a density of 3 × 10⁵ cells in a 6-cm plate for long experiments (72 h) or 1 × 10⁶ cells for short experiments (15 min–16 h). Cells were treated with either vehicle (PBS) or cET-ASOs accordingly, and protein was harvested using RIPA buffer supplemented with protease and phosphatase inhibitors (Halt Protease and Phosphatase Inhibitor Cocktail, Thermo Fisher Scientific) at the endpoint. Protein lysates were sonicated, and insoluble fractions were discarded after centrifugation at 8,900 rcf at 4°C for 5 min. Protein concentration was measured using a BCA assay (Thermo Fisher Scientific). Uniformly concentrated samples were separated by electrophoresis using NuPAGE 4–12% Bis-Tris gels (Thermo Fisher Scientific), and proteins were transferred to a nitrocellulose membrane using the Trans-Turbo transfer system (Bio-Rad). Membranes were further incubated with blocking buffer (5% BSA in TBS 0.1% Tween20) for 1 h at RT. Corresponding antibodies were incubated ON at 4°C on a rocker. Membranes were washed the following day three times with TBS-T at RT. The corresponding secondary antibodies were incubated in blocking buffer for 1 h at RT. Three washes with TBS-T were subsequently performed, and protein expression was evaluated either using the Odyssey CLx Imager (Li-Cor) or using an ECL substrate (SuperSignal West Femto substrate, Thermo Fisher Scientific) and imaging with a ChemiDoc imaging system (Bio-Rad).

KPC cells were seeded in 96-well optical bottom plates using 750 cells per well. Cells were allowed to attach and proliferate overnight. The next day, cells were treated with vehicle or the corresponding dose of cEt-ASO and allowed to proliferate for a further 72 h. At endpoint, cells were fixed with 4% paraformaldehyde solution in PBS, permeabilized using 0.1% Triton X-100 in PBS, and stained using DAPI. Plates were imaged using the Opera Phenix imaging system (Revvity) using a 5×/0.16NA objective (part HH14000402), and cell counting was performed using Harmony high-content image analysis software (Revvity). Data are expressed as proliferation index, corresponding to the fold change of cell growth between the control cells and the cET-ASO–treated cells at endpoint for each condition across independent experiments.

KPC cells were seeded in µ-Slide 8-well high slides (ibidi) using 3 × 10³ cells per well. Cells were allowed to attach and proliferate overnight. For endosomal localization experiments, cells were previously transfected in 6-cm plates using the corresponding plasmids and Lipofectamine 2000 (Thermo Fisher Scientific), and 24 h after transfection, cells were seeded into the chambered slides. The next day, cells were treated with either vehicle (PBS) or cET-ASOs accordingly and allowed to proliferate overnight (16 h). The next day, samples were fixed using a 4% paraformaldehyde solution in PBS and permeabilized with 0.1% Triton X-100 solution in PBS. Next, samples were incubated with blocking buffer (5% normal goat serum in TBS) for 1 h at RT. Primary antibodies were incubated overnight at 4°C, diluted in blocking buffer. The next day, samples were washed three times in TBS buffer and further incubated with the corresponding secondary antibodies and DAPI diluted in blocking buffer for 1 h at RT. Samples were next washed using TBS three times, and soft mounting medium (VECTASHIELD) was added to each well. Samples were visualized using either a Zeiss LSM880 microscope with Airyscan or a Zeiss Elyra 7 lattice SIM microscope, using a 63×/1.4NA objective and Zen Black software (v2.3 for Airyscan, 3.0 for Elyra). For Airyscan imaging, frame-sequential capture was used and channel-specific bandpass filters as follows: blue: 420–480 + LP605; green: 420–480 + 495–550; red: 420–480 + 495–620; far-red: 570–620 + LP645. Pixel size was 0.04 µm. Airyscan 3D processing was used at default sharpness. For SIM, lattices used were: for AlexaFluor488 G5 (27.5 µm); AlexaFluor568 G4 (32 µm) or G5 (27.5 µm) if phase modulation was judged to be sufficiently good; AlexaFluor647 G3 (36.5 µm); for DAPI G6 (23 µm) with 13 phases. z-stacks were captured at intervals of 91–125 nm depending on the experiment, or for some samples at 55 nm (over-sampling). Lasers (488, 561 and 642 nm, all 500 mW; 405 nm 50 mW) were typically used at 2.5–4% power. Emitted light was captured on PCO edge 4.2M sCMOS cameras via a Duolink adaptor, using a dual band-pass filter (490–560 + LP640) with a 50-ms exposure time. Processing was by SIM² using “standard live” settings of input SNR medium, iterations 16, regularization 0.065; input and output sampling was set to ×4, median filter. Grating period was 718.33 nm, and the resulting xy scaling was 0.031 µm. Image analysis was performed using Fiji-ImageJ software to obtain Pearson correlation coefficients and fluorescence intensity profiles of representative vesicle examples. Fluorescence intensity profiles are shown as the fold change relative to the maximum intensity value for each fluorophore. For 3D image reconstruction, Imaris software was used to create the 3D surfaces using the SIM² data generated after processing the images obtained using the Elyra7 microscope.

Opera Phenix (Revvity) and Harmony v5.2 software were used to acquire high-content confocal imaging. KPC cells were seeded on 96-well optical plates (Revvity Phenoplate, 5,000 cells/well). After 24 h, cells were treated with vehicle (PBS) or cET-ASOs as appropriate and allowed to proliferate overnight (for the 16-h time points). For the 4-h time points, cells were treated with vehicle (PBS) or cET-ASOs the following day. Vehicle and cET-ASOs were combined with bafilomycinA1 or chloroquine treatments at the appropriate concentrations. Cells were then fixed, permeabilized, and stained using the corresponding primary antibodies overnight at 4 °C. The following day, cells were incubated with DAPI and the corresponding fluorescent secondary antibodies at RT for 1 h. Cells were then imaged in an Opera Phenix (Revvity) high-throughput confocal microscope using either 20×/1.0NA water immersion (part HH14000421) or 63×/1.15NA water immersion (part HH14000423) objectives, and images were further analyzed using Harmony v5.2 software. For distance analysis, for each cell, vesicle distance to the nucleus was measured, and cells were binned into proximal groups (0–1 µm) and distal groups (from 1 µm to the maximum value). Then the mean fluorescence value for each particle was assessed for the corresponding markers such as EphA2, galectin-9, or ASO. Graphical representations and statistical analysis were performed using R version 4.5.0 within RStudio, running a custom pipeline for importing, harmonizing, and analyzing Opera Phenix/Harmony high-content imaging data. Data preprocessing relied on dplyr (1.1.4), readr (2.1.6), stringr (1.6.0), forcats (1.0.1), rlang (1.1.7), and tidyr. Visualization was performed using ggplot2 (4.0.1), ggridges (0.5.7), patchwork (1.3.2), and the HCL-based palette tools from colorspace (2.1–2). Interactive file selection and RStudio integration used rstudioapi (0.18.0). Statistical analysis modules used rstatix (0.7.3) for baseline tests, while more advanced models and contrasts relied on emmeans, lme4, lmerTest, and multcomp.

These approaches were adapted from previously established protocols (Bhattacharya et al., 2017). 1 mg CD44 (Sino Biologicals) was dissolved in PBS to a concentration of 200 µM, from which a serial dilution was generated in PBS. For binding experiments, 50 nM Cy3-ASO or 1 μM Cy3-HA were incubated with increasing concentrations of CD44 (454 nM–50 μM) in PBS for 1 h at 30°C in a total volume of 10 μl. For competition experiments, 50 nM Cy3-ASO was incubated with 30 μM CD44 for 30 min at RT before increasing concentrations of unlabelled competitor HA (454 nM–50 μM) were added to the reaction. Fluorescence polarization of samples was then measured in a Tecan Spark (λ_ex = 535 nm; λ_em > 595 nm). Data were analyzed using GraphPad Prism (v10.5). To derive the dissociation constants, a 1:1 binding model was used to fit the data. Experiments were performed in triplicate.

Proximity ligation was performed following the manufacturer’s (Merck Sigma-Aldrich) instructions. Briefly, 10,000 cells were seeded on an IBIDI 8-well chamber slide and allowed to attach overnight. The following day, cEt-ASOs were added to generate the 16-h time point and incubated overnight. 24 h later, cEt-ASOs were added to generate the 4-h time point. Cells were fixed in 4% paraformaldehyde for 15 min at RT and permeabilized using 0.3% Triton X-100 in PBS for 7 min. Cells were blocked using the blocking solution according to the manufacturer’s instructions, and the primary CD44-minus probe and the ASO antibody were mixed and incubated overnight at 4°C. Cells were incubated with the PLUS PLA probes in conjunction with Phalloidin-488 (1:400) and DAPI for 1 h at 37°C. Ligation and amplification protocols were then performed following the manufacturer’s instructions, and cells were prepared for imaging. Imaging was performed using a Zeiss LSM880 microscope incorporating Airyscan as detailed above. Images were analyzed using Fiji/ImageJ, and graphical representation and statistical analysis were performed using RStudio software as detailed above.

Opera Phenix (Revvity) and Harmony 5.2 software were used to acquire high-content confocal live imaging. KPC cells were seeded on 96-well optical plates (Revvity Phenoplate, 5,000 cells/well). 24 h later, cells were treated with vehicle (PBS) or cET-ASOs accordingly and allowed to proliferate overnight (16 h), in combination with either vehicle or liproxstatin-1 (1 μM). The next day, for the 4-h time points, cells were treated with vehicle (PBS) or cET-ASOs accordingly, as well as with either vehicle or liproxstatin-1. Cells were then stained with either CellROX green (5 μM) for total ROS quantification or C11-bodipy (5 μM) to assess LP, in combination with Hoechst 33342 to stain nuclei for 45 min at 37°C. Cells were then imaged in an Opera Phenix high-throughput confocal microscope using a 20×/1.0NA water immersion objective (part HH14000421) at 37°C and 5% CO₂. Z-stacks were acquired and further analyzed using Harmony v5.2 software. Since CellROX green oxidation promotes its translocation to the nucleus and DNA binding (see the manufacturer’s CellROX Oxidative Stress Reagents manual), ROS production was measured as the ratio of CellROX detected in the nucleus versus the amount present in the cytoplasm. For C11-bodipy, LP was calculated as the fluorescence intensity of oxidized lipids (green, 488/510 nm) expressed as a ratio of that of reduced state lipids (red, 581/591 nm). Graphs and statistical analyses were performed using R (version 4.5.0) within RStudio, running a custom pipeline for importing, harmonizing, and analyzing Opera Phenix/Harmony high-content imaging data. Data preprocessing, visualization and statistical analysis modules were used as described above. High-resolution confocal images of KPC cells incubated with the C11-bodipy probe were obtained using a Zeiss LSM880 microscope comprising Airyscan and an incubation chamber at 37°C and 5% CO₂.

H1299 cells were transfected with SR siRNA pools (see details in the “Reagents” section) using Nucleofection Kit V and an AMAXA nucleofector (Lonza). 1.5 × 10⁵ cells were then seeded onto 6-cm dishes and allowed to attach overnight. The next day, cells were incubated for 72 h with the corresponding doses of cEt-ASO. Cells were then lysed, and either protein or RNA was harvested. Effectiveness of siRNA knockdown as well as KRAS mRNA targeting and downstream signalling by cEt-ASOs were assessed by either western blotting or real-time PCR following the protocols described above.

Suppression of specific gene expression and generation of stable cell lines were achieved using the CRISPR-knockout genome editing system. Cloning of guide RNAs into lentiviral plasmids and generation of stable cell lines after lentiviral infection were performed according to the protocol established by the Zhang lab (Ran et al., 2013). Briefly, guide RNAs (see details in the “Reagents” section) were designed, annealed, and inserted into lentiCRISPRv2 lentiviral plasmids (#52961; Addgene). Lentiviral particles were produced by transducing HEK293FT cells with the lentiCRISPRv2 guide RNA–expressing plasmid together with psPAX2 and VSV-G packaging plasmids. Supernatants were collected after 48 h and transferred onto the recipient cells. Selection of positively transduced cells was assessed with antibiotic selection (puromycin) in the culture medium. After three passages, the successful knock-out of the target genes was assessed by qPCR as well as by western blotting.

KPC cells stably expressing an EPHA2-TurboID construct (Marco et al., 2021) were seeded in 10-cm plates using 2 × 10⁶ cells per plate. Cells were allowed to attach and proliferate overnight. The next day, cells were incubated with 100 µM biotin and 5 µM cET-ASO for 16 h. The next day, cells were placed on ice to stop the biotin ligation reaction and washed five times with ice-cold PBS. Cells were scraped from the plate in PBS and pelleted at 300×g for 3 min at 4°C. Cells were lysed in RIPA buffer supplemented with protease inhibitors and sonicated. The insoluble fraction was discarded, and the free biotin excess from the lysates was removed using an AMICON 3K filter column (Millipore). Cell lysates were recovered and further incubated with streptavidin-agarose beads (Millipore) for 2 h at 4°C. After the incubation, the beads were further washed sequentially with RIPA, 1M KCl, 0.1M Na₂CO₃, 2M urea, and a final 100 mM NaCl—10 mM Tris, pH 7.5—0.5 mM EDTA buffer. Beads were then resuspended in loading buffer and 2 mM biotin, and biotinylated targets were assessed using western blotting.

For immunofluorescence, KPC cells stably expressing an EPHA2-TurboID construct were plated onto glass-bottom 35-mm dishes and allowed to attach overnight. The next day, cells were treated with biotin as described above, and 16 h afterward, biotin was washed five times with PBS, and the cells were fixed using 4% paraformaldehyde. Cells were processed as described above, and biotinylated proteins were assessed using streptavidin conjugated to Alexa488 (Thermo Fisher Scientific). Samples were visualized in a Zeiss880 Airyscan microscope as described above.

Publicly available data, analysis, and graphs were obtained from cBioPortal (Cerami et al., 2012) using the TCGA dataset corresponding to PAAD. GEPIA2 portal (Tang et al., 2019) was used to obtain the comparison between tumor and normal tissue data. Further genomic data regarding mRNA expression in PDAC were obtained from analysis of microarray and RNA-seq data by Bailey et al. (2016), as well as PDAC subtype classification expression data.

Reagents or resources are shown in Table 1.

Fig. S1 shows supplementary data relates to main Figs. 1 and 2. Fig. S1, A–H inclusive, presents data pertaining to human PDAC supporting main Fig. 1, A–C. Fig. S1, I–Q inclusive, presents data supporting the analysis of the influence of cET-ASO^Kras on KPC and H1299 cells. Fig. S2 shows supplementary data relates to main Fig. 3 and presents data supporting the analysis of SR function and the signalling downstream of these receptors in ASO uptake. Fig. S3 shows supplementary data relates to main Fig. 5 and to the more discursive, final section of the results and discussion. Fig. S3, A–M inclusive, presents data supporting the analysis of endosome leakiness and the role of SG components in endosomal repair covered in Fig. 5. Fig. S3 N presents a schematic summary depicting the endocytic pathway transporting ASOs from the extracellular space into the perinuclear region and enabling their release from nuclear-captured leaky endosomes.

The data supporting the findings of this study are available within the article and its supplementary information files and from the corresponding author upon request.

We would like to thank core staff in the Biological Services Unit, the Beatson Advanced Imaging Resource (RRID:SCR_023875), Molecular Technology Services (RRID:SCR_027368), and Histology Facility and Central Services (CRUK Scotland Institute) for their support, which facilitated the work described in this paper. The results shown here are in whole or part based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. The data used for the analyses described in this manuscript were obtained from the GTEx Portal and/or dbGaP accession number phs000424.vN.pN. This paper was critically reviewed by Catherine Winchester (CRUK Scotland Institute; RRID:SCR_027384).

This work was funded by a Cancer Research UK Core Programme Award to J.C. Norman (A18277), Breast Cancer Now (2018NovPR1268), and a Medical Research Council project grant to M. Bushell and J.C. Norman (MR/P01058×/1). We acknowledge the Cancer Research UK Glasgow Centre (C596/A18076) and the BSU facilities at the Cancer Research UK Scotland Institute (C596/A17196 and A31287).

Author contributions: Sergi Marco: conceptualization, data curation, formal analysis, investigation, methodology, project administration, supervision, visualization, and writing—original draft, review, and editing. Peter J. Walsh: conceptualization, investigation, and methodology. Alexey S. Revenko: methodology, resources, and writing—review and editing. Tobias Schmidt: conceptualization, formal analysis, investigation, methodology, supervision, and visualization. Peter A. Thomason: investigation and writing—review and editing. Lynn McGarry: formal analysis. A. Robert MacLeod: investigation and writing—review and editing. Sonam Ansel: investigation. Dina Tataran: validation. Martin Bushell: conceptualization. Chiara Braconi: funding acquisition, resources, and writing—review and editing. Jim C Norman: conceptualization, funding acquisition, investigation, methodology, project administration, supervision, visualization, and writing—original draft, review, and editing.